Soket AI Secures 177 Crore for Indic Language AI Development

Soket AI, an Indian technology startup, aims to bridge the language gap in artificial intelligence by developing large language models (LLMs) that prioritize Indian languages and other under-represented tongues globally. The company has secured ₹177 crore (approximately $2.1 million USD) in funding from the IndiaAI Mission, a significant portion of which is allocated to compute resources.

Founded in 2019, Soket AI is building a sovereign, open-source AI stack designed to serve a range of Indian languages. Its approach distinguishes itself by not only focusing on multilingual capabilities but also on developing robust reasoning, coding, and problem-solving skills applicable across various languages. The goal is to reduce reliance on foreign AI providers and establish an indigenous AI system.

A key initiative for Soket AI is "Project EKA," an ambitious effort to create a 120-billion parameter, open-source AI model. Planned for launch in 2027, this model intends to support over 20 languages across the Global South, addressing a critical need for digital representation in regions with limited high-quality datasets. The company is investing heavily in mathematics, coding, and structured logic, which it considers foundational pillars of advanced AI.

Soket AI has encountered challenges in generating high-quality datasets, particularly for Indian and low-resource languages. To overcome this, the startup has developed proprietary pipelines for data collection, cleaning, and structuring. Furthermore, it is investing in "harness engineering," essentially orchestration infrastructure, to ensure reliable AI deployment in production and seamless integration with enterprise applications.

As its operations expand, Soket AI is actively recruiting specialized researchers and engineers. Funding for its growth comes from the IndiaAI Mission and new capital investments, as the company seeks to further its ambitious goals in AI research and development.